the Mai/netic Effect of Electric Convection. 445 



from the disks to the condensing plates, the magnetic effect 

 observed fell practically to zero; whereas the potential of 

 the disk, as measured by an electrometer connected thereto, 

 showed a decrease of less than yo of the value observed 

 just before the sparks began. In Cremieu^'s apparatus it 

 was impossible to see such sparks if they should occur, as 

 the disk turned in a completely closed box of cast iron. 

 However, in spite of the fact that the disk and condensing 

 plates were covered with a thin layer of caoutchouc, it was 

 possible that such sparks might occur, as the distance 

 between these plates was very small. 



Only a galvanometric measurement of the quantity of 

 electricity taken by the disk at each charge, or the quantity 

 given up at each discharge, could decide this point. AYe 

 made these measurements, employing voltages from 1000 to 

 5500 volts. For 1000 volts, it is certain that no spark can 

 occur across o'b mm, of air. As the voltage is increased, 

 one should obtain currents j^roportional to the voltage as 

 long as no sparks occur. If such sparks take place, the 

 charging current should increase more rapidly, the discharge 

 current less rapidly, than the voltage. Measurements showed 

 that up to 5500 volts, the maximum potential employed. 

 there was a strict proportionality between the charge and 

 voltage. 



The cause of Cremieu's negative results was not then in 

 this. 



Verifications in Pender's Experiments. — In the series of 

 experiments on open currents made by Cremieu * in 1 902, 

 with the aid of M. J. Javal, certain peculiar magnetic 

 effects were observed in the neighbourhood of a node of 

 electrical oscillations, such as is formed by the turning disk 

 in the convection experiments. The conditions for the 

 production of these effects are so far similar to the con- 

 ditions in Pender's experiment that one might believe that 

 the effects observed by Pender were due to such oscillations, 

 and not to convection itself. 



This hypoihesis was rendered still more plausible by 

 certain magnetic effects observed wdien the disks are charged 

 or discharged at rest. These effects, though extremely irre- 

 gular, are capable of giving to the galvanometer connected 

 with the induced coil deflexions of more than 100 mm. 

 Moreover, these eff'ects are considerably augmented by the 

 smallest hole in the electric screen of tinfoil w^hich protects 

 the coil. Again, as the tinfoil is wrapped around the 

 winding of the coil only, thus leaving the centre of the coil 



* Comptes Renclus, vol. cxxxv. p. 27 (1902), and Joia-n, de Phys. 

 Dec. 1902. 



